Useless Project

VadaScope is an over-engineered AI that settles the age-old debate of what makes a perfect vada. Using a sophisticated CycleGAN model, our system analyzes an image of a vada and provides an objective, unsolicited Vada Perfection Index (VPI) score, ending family arguments one snack at a time.

Type Competition

Role Competitor

Timeline 2025

Link

VadaScope [VPI- Vada Perfection Index]🎯

Basic Details

Team Name: Super Nova

Team Members

-Team Lead: Sreehari R - Sree Buddha College Of Engineering, Pattoor -Member 2:  Abhinav R - Sree Buddha College Of Engineering, Pattoor

Project Description

VadaScope is an over-engineered AI that settles the age-old debate of what makes a perfect vada. Using a sophisticated CycleGAN model, our system analyzes an image of a vada and provides an objective, unsolicited Vada Perfection Index (VPI) score, ending family arguments one snack at a time.

The Problem (that doesn’t exist)

Every day, millions of people consume vadas of questionable quality. The lack of a universal, quantifiable metric for vada perfection—from its roundness to its hole-to-vada ratio—threatens the very fabric of our tea-time traditions. This rampant subjectivity is a recipe for culinary chaos and inconsistent snacking experiences.

The Solution (that nobody asked for)

Our solution is a two-stage hybrid pipeline that combines the predictive power of deep learning with the analytical precision of classical computer vision to deliver the ultimate Vada Perfection Index (VPI) score.

✨Stage 1: The AI Oracle (CycleGAN) We start by feeding a real-world vada image into a trained CycleGAN model. Acting as an AI Oracle, the model “hallucinates” idealized markings onto the input image — essentially generating the platonic ideal of what the vada should look like.

The result? A marked image representing the Vada Perfection that could have been.

✨Stage 2: The Scrutinizer (CV + VPI Calculation) The marked image is then passed through a classical Computer Vision module, which extracts four key metrics. These metrics are used to compute the final Vada Perfection Index (VPI_S) as follows:

VPI_S = (w_size · S_size + w_shape · S_shape + w_hole · S_hole + w_color · S_color) × 100

📏 Metric Breakdown

🎚️Weight Calibration w_size  = 0.01   w_shape = 0.40   w_hole  = 0.30   w_color = 0.29

Technical Details

Technologies/Components Used

For Software:

-Languages used: Python -Frameworks used: PyTorch, Flask -Libraries used: - OpenCV (opencv-python) for computer vision analysis - Pillow (PIL) for image manipulation and report generation - Albumentations for data augmentation - NumPy for numerical operations -Tools used: - Git / GitHub for version control - Visual Studio Code

For Hardware:

-Main components: A PC/Laptop with a dedicated NVIDIA GPU. A camera (e.g., phone camera) to capture images of vadas. -Specifications: A CUDA-enabled NVIDIA GPU is required for model training (e.g., NVIDIA GeForce RTX 4050 Laptop GPU). -Tools required: Not applicable.

Project Documentation

For Software:

Screenshots

Diagrams

Project Demo

Video

https://github.com/user-attachments/assets/0d0acf19-0593-4760-ae34-c46289d867f2

Additional Demos

https://github.com/user-attachments/assets/d4f79607-52d5-4537-8e83-4a740e043168

Team Contributions

-Sreehari R: Led the backend development, including setting up the Flask server, training the CycleGAN model, and implementing the computer vision logic for VPI calculation. -Abhinav R: Designed and developed the main conceptual idea of the project and the frontend user interface, including the HTML structure, all CSS styling and animations, and the JavaScript logic for handling user interaction and displaying results.

Made with ❤️ at TinkerHub Useless Projects

other works